Listening Forward: Next Patch Embedding Prediction Enables Scalable Audio Learners

Hugging Face Daily Papers Papers

Summary

NAPE is a self-supervised audio learning framework that uses causal Transformers to predict next spectrogram patch embeddings, achieving state-of-the-art performance on multiple audio and speech benchmarks with a minimalist design.

Self-supervised learning (SSL) has driven substantial progress in audio representation learning, though existing methods have increasingly relied on elaborate pre-training recipes to reach competitive performance. A markedly different pre-training philosophy underpins the most influential progress in language modeling and, more recently, in visual representation learning: rather than train encoders as static feature extractors, models are trained to predict the next element, a discrete token or a continuous embedding, from the preceding context. Autoregressive prediction thereby provides a unified pre-training interface that transfers across modalities, compelling the model to learn the underlying data distribution. We ask whether such a simple causal paradigm can yield strong audio learners, given that audio's temporal structure makes autoregressive prediction of patch embeddings a natural fit. We introduce NAPE (Next-Audio-Patch-Embedding prediction), a self-supervised framework in which a causal Transformer predicts each next patch embedding of a log-mel spectrogram from the previous ones, using causal masking and stop-gradient as its sole training signal. The design is intentionally minimalist, avoiding reconstruction decoders, acoustic tokenizers, student-teacher setups, and auxiliary regularization losses. Across six audio and speech benchmarks, NAPE achieves state-of-the-art fine-tuning performance on several tasks, scales consistently across encoder sizes, and yields strong linear-probing results. NAPE also produces structured attention patterns without explicit supervision.
Original Article
View Cached Full Text

Cached at: 08/21/26, 08:08 AM

Paper page - Listening Forward: Next Patch Embedding Prediction Enables Scalable Audio Learners

Source: https://huggingface.co/papers/2608.19863

Abstract

NAPE uses causal Transformers to predict successive spectrogram patch embeddings for self-supervised audio learning without auxiliary components.

Self-supervised learning(SSL) has driven substantial progress in audio representation learning, though existing methods have increasingly relied on elaborate pre-training recipes to reach competitive performance. A markedly different pre-training philosophy underpins the most influential progress in language modeling and, more recently, in visual representation learning: rather than train encoders as static feature extractors, models are trained to predict the next element, a discrete token or a continuous embedding, from the preceding context.Autoregressive predictionthereby provides a unified pre-training interface that transfers across modalities, compelling the model to learn the underlying data distribution. We ask whether such a simple causal paradigm can yield strong audio learners, given that audio’s temporal structure makesautoregressive predictionofpatch embeddingsa natural fit. We introduce NAPE (Next-Audio-Patch-Embedding prediction), a self-supervised framework in which acausal Transformerpredicts each next patch embedding of alog-mel spectrogramfrom the previous ones, usingcausal maskingandstop-gradientas its sole training signal. The design is intentionally minimalist, avoiding reconstruction decoders, acoustic tokenizers, student-teacher setups, and auxiliary regularization losses. Across six audio and speech benchmarks, NAPE achieves state-of-the-art fine-tuning performance on several tasks, scales consistently across encoder sizes, and yields strong linear-probing results. NAPE also produces structured attention patterns without explicit supervision.

View arXiv pageView PDFProject pageGitHubAdd to collection

Get this paper in your agent:

hf papers read 2608\.19863

Don’t have the latest CLI?curl \-LsSf https://hf\.co/cli/install\.sh \| bash

Models citing this paper0

No model linking this paper

Cite arxiv.org/abs/2608.19863 in a model README.md to link it from this page.

Datasets citing this paper0

No dataset linking this paper

Cite arxiv.org/abs/2608.19863 in a dataset README.md to link it from this page.

Spaces citing this paper0

No Space linking this paper

Cite arxiv.org/abs/2608.19863 in a Space README.md to link it from this page.

Collections including this paper0

No Collection including this paper

Add this paper to acollectionto link it from this page.

Similar Articles

Introducing next-generation audio models in the API

OpenAI Blog

OpenAI introduced next-generation audio models for the API, including improved speech-to-text (gpt-4o-transcribe, gpt-4o-mini-transcribe) and customizable text-to-speech models that enable developers to build more intelligent and expressive voice agents with enhanced accuracy across challenging scenarios.

AudioMosaic: Contrastive Masked Audio Representation Learning

arXiv cs.LG

AudioMosaic introduces a contrastive learning-based audio encoder that uses structured time-frequency masking on spectrogram patches for efficient large-batch training, achieving state-of-the-art performance on audio benchmarks and improving audio-language models.